GTFD‐XTNet: A tabular learning‐based ensemble approach for short‐term prediction of photovoltaic power
Year: 2022
Type: article
Abstract: Greenhouse effect has led to the deterioration of the global climate, so the development of renewable energy is extremely urgent. Photovoltaic (PV) generation has attracted much attention in recent years, but it is limited by the uncertain and intermittent nature, which needs promising PV prediction models. In this study, we propose a novel hybrid deep learning model based on the TabNet, extreme gradient boosting (XGBoost) and linear regression, which is motivated by the ensemble and residual ideas. The models of XGBoost and TabNet are fitted simultaneously based on numeric weather prediction (NWP) information. We adopt the linear regression to combine the results from both TabNet and XGBoost to render the final estimates through ResNet approaches. Besides, the gradient boosting decision tree (GBDT) is incorporated for redundant feature reduction and important feature selection. Meanwhile, the general time features difference (GTFD) is also developed in this work to augment the original feature space and to mine the implicit temporal relationships among features. The proposed model is validated in real‐life PV data and the experimental results show that the GTFD‐XTNet model outperforms the state‐of‐the‐art forecasting models in the PV field.
Source: IET Renewable Power Generation
DOI: 10.1049/rpg2.12494
Authors: Qinglun Wang, Songjian Chai, Yun Liu, Guibin Wang
Institutions: Shenzhen University, South China University of Technology
Language: English
Field: Energy
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,300
Award: G1125668645
OpenAlex ID: W4285386802
